FeaSc: A framework for single cell pathway analysis

Introduction

We previously developed PaaSc, a method for inferring pathway activity from single-cell and spatial transcriptomics data. PaaSc employs multiple correspondence analysis (MCA) to simultaneously project cells and genes into a common latent space, then selects pathway-associated dimensions through linear regression to infer pathway activity scores. We validated PaaSc across diverse benchmarking datasets, including those with joint protein and RNA profiling, as well as large-scale cancer scRNA-seq cohorts. Compared with state-of-the-art methods, PaaSc demonstrated superior performance across multiple applications: scoring cell type-specific gene sets, identifying cell senescence-associated pathways, and exploring GWAS trait-associated cell types. Importantly, PaaSc maintained accuracy despite batch effects and showed robust performance across different data modalities, including scATAC-seq and spatial transcriptomics data. Despite these strengths, PaaSc had several limitations: it was implemented only in R and restricted to MCA-based dimensionality reduction. To address these constraints and expand functionality, we developed FeaSc, a Python package for inferring pathway activity from single-cell and spatial transcriptomic data. Feasc offers three key improvements: (i) it supports multiple dimensionality reduction methods beyond MCA, including PCA and NMF; (ii) it can infer signaling pathway activity (such as cytokine signaling) using ridge regression when pathways are defined as gene expression changes rather than gene sets; and (iii) for datasets with batch effects, Feasc integrates batch-corrected data from scVI for more accurate pathway activity inference.

Installation

Documentation

  1. Demo dataset tutorial
  2. Pathway activity inference for one sample tutorial
  3. Pathway activity inference for multiple samples tutorial
  4. Prediction of cytokine signaling activity tutorial

Benchmarking

Datasets